[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
[avr-libc-dev] poor optimisation
|
From: |
Rob Ward |
|
Subject: |
[avr-libc-dev] poor optimisation |
|
Date: |
Mon, 18 Nov 2002 10:31:03 +1300 |
|
User-agent: |
Mozilla/5.0 (Windows; U; Windows NT 5.0; en-US; rv:1.1) Gecko/20020826 |
Hi. I think I have an example of some very poor avr-gcc optimisation. I
have pasted my c code below, followed by a .lst listing. I have also
included my makefile at the end.
c code ...
SIGNAL(SIG_OVERFLOW0) { // signal handler for tcnt0
overflow interrupt - SIGNAL disables interrupts
u08 DataHigh, DataMid, DataLow; // 3 bytes to hold 19 bits to
send to 3972
outp(Timer0Interval, TCNT0);
if (InConfigMode == TRUE) { // At startup, we are
InConfigMode so we can send the 19 config bits.
InConfigMode = FALSE; // Enter here only once at startup
DataHigh = 0x80;
DataMid = 0x30;
DataLow = 0x01;
}
else {
if (StepDirn == CCW) {
DataHigh = PRG_RDB(&WordHighCCW[StepPos]);
DataMid = PRG_RDB(&WordMidCCW[StepPos]);
DataLow = PRG_RDB(&WordLowCCW[StepPos]);
}
else { // step direction is clockwise
DataHigh = PRG_RDB(&WordHighCW[StepPos]);
DataMid = PRG_RDB(&WordMidCW[StepPos]);
DataLow = PRG_RDB(&WordLowCW[StepPos]);
}
}
SetStrobeLow(); // set strobe low to initiate
data transfer
if (DataHigh & 0x80) SetDataHigh(); else SetDataLow();
SetClockHigh(); SetClockLow();
if (DataHigh & 0x40) SetDataHigh(); else SetDataLow();
SetClockHigh(); SetClockLow();
if (DataHigh & 0x20) SetDataHigh(); else SetDataLow();
SetClockHigh(); SetClockLow();
if (DataHigh & 0x10) SetDataHigh(); else SetDataLow();
SetClockHigh(); SetClockLow();
if (DataHigh & 0x08) SetDataHigh(); else SetDataLow();
SetClockHigh(); SetClockLow();
if (DataHigh & 0x04) SetDataHigh(); else SetDataLow();
SetClockHigh(); SetClockLow();
if (DataHigh & 0x02) SetDataHigh(); else SetDataLow();
SetClockHigh(); SetClockLow();
if (DataHigh & 0x01) SetDataHigh(); else SetDataLow();
SetClockHigh(); SetClockLow();
if (DataMid & 0x80) SetDataHigh(); else SetDataLow();
SetClockHigh(); SetClockLow();
if (DataMid & 0x40) SetDataHigh(); else SetDataLow();
SetClockHigh(); SetClockLow();
if (DataMid & 0x20) SetDataHigh(); else SetDataLow();
SetClockHigh(); SetClockLow();
if (DataMid & 0x10) SetDataHigh(); else SetDataLow();
SetClockHigh(); SetClockLow();
if (DataMid & 0x08) SetDataHigh(); else SetDataLow();
SetClockHigh(); SetClockLow();
if (DataMid & 0x04) SetDataHigh(); else SetDataLow();
SetClockHigh(); SetClockLow();
if (DataMid & 0x02) SetDataHigh(); else SetDataLow();
SetClockHigh(); SetClockLow();
if (DataMid & 0x01) SetDataHigh(); else SetDataLow();
SetClockHigh(); SetClockLow();
if (DataLow & 0x04) SetDataHigh(); else SetDataLow();
SetClockHigh(); SetClockLow();
if (DataLow & 0x02) SetDataHigh(); else SetDataLow();
SetClockHigh(); SetClockLow();
if (DataLow & 0x01) SetDataHigh(); else SetDataLow();
SetClockHigh(); SetClockLow();
SetStrobeHigh(); // reset strobe to high to
terminate data transfer
StepCompleted = TRUE;
}
and here is the .lst output
note all the (lots of) wasted rom space because of these...
14a: c0 98 cbi 0x18, 0 ; 24
14c: c3 cf rjmp .-122 ; 0xd4
///////////////////////////////////////////////////////////////////
SIGNAL(SIG_OVERFLOW0) { // signal handler for tcnt0 overflow
interrupt - SIGNAL disables interrupts
0: 1f 92 push r1
2: 0f 92 push r0
4: 0f b6 in r0, 0x3f ; 63
6: 0f 92 push r0
8: 11 24 eor r1, r1
a: 2f 93 push r18
c: 3f 93 push r19
e: 8f 93 push r24
10: 9f 93 push r25
12: ef 93 push r30
14: ff 93 push r31
u08 DataHigh, DataMid, DataLow; // 3 bytes to hold 19 bits to send to
3972
outp(Timer0Interval, TCNT0);
16: 80 91 00 00 lds r24, 0x0000
1a: 82 bf out 0x32, r24 ; 50
if (InConfigMode == TRUE) { // At startup, we are InConfigMode so we
can send the 19 config bits.
1c: 80 91 00 00 lds r24, 0x0000
20: 81 30 cpi r24, 0x01 ; 1
22: 09 f4 brne .+2 ; 0x26
24: c1 c0 rjmp .+386 ; 0x1a8
InConfigMode = FALSE; // Enter here only once at startup
DataHigh = 0x80;
DataMid = 0x30;
DataLow = 0x01;
}
else {
if (StepDirn == CCW) {
26: 80 91 00 00 lds r24, 0x0000
2a: 81 30 cpi r24, 0x01 ; 1
2c: 09 f4 brne .+2 ; 0x30
2e: a3 c0 rjmp .+326 ; 0x176
DataHigh = PRG_RDB(&WordHighCCW[StepPos]);
DataMid = PRG_RDB(&WordMidCCW[StepPos]);
DataLow = PRG_RDB(&WordLowCCW[StepPos]);
}
else { // step direction is clockwise
DataHigh = PRG_RDB(&WordHighCW[StepPos]);
30: 80 91 00 00 lds r24, 0x0000
34: e8 2f mov r30, r24
36: ff 27 eor r31, r31
38: e0 50 subi r30, 0x00 ; 0
3a: f0 40 sbci r31, 0x00 ; 0
3c: c8 95 lpm
3e: 90 2d mov r25, r0
DataMid = PRG_RDB(&WordMidCW[StepPos]);
40: 80 91 00 00 lds r24, 0x0000
44: e8 2f mov r30, r24
46: ff 27 eor r31, r31
48: e0 50 subi r30, 0x00 ; 0
4a: f0 40 sbci r31, 0x00 ; 0
4c: c8 95 lpm
4e: 20 2d mov r18, r0
DataLow = PRG_RDB(&WordLowCW[StepPos]);
50: 80 91 00 00 lds r24, 0x0000
54: e8 2f mov r30, r24
56: ff 27 eor r31, r31
58: e0 50 subi r30, 0x00 ; 0
5a: f0 40 sbci r31, 0x00 ; 0
5c: c8 95 lpm
5e: 30 2d mov r19, r0
}
}
SetStrobeLow(); // set strobe low to initiate data
transfer
60: 96 98 cbi 0x12, 6 ; 18
if (DataHigh & 0x80) SetDataHigh(); else SetDataLow(); SetClockHigh();
SetClockLow();
62: 97 ff sbrs r25, 7
64: 86 c0 rjmp .+268 ; 0x172
66: c0 9a sbi 0x18, 0 ; 24
68: 97 9a sbi 0x12, 7 ; 18
6a: 97 98 cbi 0x12, 7 ; 18
if (DataHigh & 0x40) SetDataHigh(); else SetDataLow(); SetClockHigh();
SetClockLow();
6c: 89 2f mov r24, r25
6e: 99 27 eor r25, r25
70: 86 ff sbrs r24, 6
72: 7d c0 rjmp .+250 ; 0x16e
74: c0 9a sbi 0x18, 0 ; 24
76: 97 9a sbi 0x12, 7 ; 18
78: 97 98 cbi 0x12, 7 ; 18
if (DataHigh & 0x20) SetDataHigh(); else SetDataLow(); SetClockHigh();
SetClockLow();
7a: 85 ff sbrs r24, 5
7c: 76 c0 rjmp .+236 ; 0x16a
7e: c0 9a sbi 0x18, 0 ; 24
80: 97 9a sbi 0x12, 7 ; 18
82: 97 98 cbi 0x12, 7 ; 18
if (DataHigh & 0x10) SetDataHigh(); else SetDataLow(); SetClockHigh();
SetClockLow();
84: 84 ff sbrs r24, 4
86: 6f c0 rjmp .+222 ; 0x166
88: c0 9a sbi 0x18, 0 ; 24
8a: 97 9a sbi 0x12, 7 ; 18
8c: 97 98 cbi 0x12, 7 ; 18
if (DataHigh & 0x08) SetDataHigh(); else SetDataLow(); SetClockHigh();
SetClockLow();
8e: 83 ff sbrs r24, 3
90: 68 c0 rjmp .+208 ; 0x162
92: c0 9a sbi 0x18, 0 ; 24
94: 97 9a sbi 0x12, 7 ; 18
96: 97 98 cbi 0x12, 7 ; 18
if (DataHigh & 0x04) SetDataHigh(); else SetDataLow(); SetClockHigh();
SetClockLow();
98: 82 ff sbrs r24, 2
9a: 61 c0 rjmp .+194 ; 0x15e
9c: c0 9a sbi 0x18, 0 ; 24
9e: 97 9a sbi 0x12, 7 ; 18
a0: 97 98 cbi 0x12, 7 ; 18
if (DataHigh & 0x02) SetDataHigh(); else SetDataLow(); SetClockHigh();
SetClockLow();
a2: 81 ff sbrs r24, 1
a4: 5a c0 rjmp .+180 ; 0x15a
a6: c0 9a sbi 0x18, 0 ; 24
a8: 97 9a sbi 0x12, 7 ; 18
aa: 97 98 cbi 0x12, 7 ; 18
if (DataHigh & 0x01) SetDataHigh(); else SetDataLow(); SetClockHigh();
SetClockLow();
ac: 80 ff sbrs r24, 0
ae: 53 c0 rjmp .+166 ; 0x156
b0: c0 9a sbi 0x18, 0 ; 24
b2: 97 9a sbi 0x12, 7 ; 18
b4: 97 98 cbi 0x12, 7 ; 18
if (DataMid & 0x80) SetDataHigh(); else SetDataLow(); SetClockHigh();
SetClockLow();
b6: 27 ff sbrs r18, 7
b8: 4c c0 rjmp .+152 ; 0x152
ba: c0 9a sbi 0x18, 0 ; 24
bc: 97 9a sbi 0x12, 7 ; 18
be: 97 98 cbi 0x12, 7 ; 18
if (DataMid & 0x40) SetDataHigh(); else SetDataLow(); SetClockHigh();
SetClockLow();
c0: 82 2f mov r24, r18
c2: 99 27 eor r25, r25
c4: 86 ff sbrs r24, 6
c6: 43 c0 rjmp .+134 ; 0x14e
c8: c0 9a sbi 0x18, 0 ; 24
ca: 97 9a sbi 0x12, 7 ; 18
cc: 97 98 cbi 0x12, 7 ; 18
if (DataMid & 0x20) SetDataHigh(); else SetDataLow(); SetClockHigh();
SetClockLow();
ce: 85 ff sbrs r24, 5
d0: 3c c0 rjmp .+120 ; 0x14a
d2: c0 9a sbi 0x18, 0 ; 24
d4: 97 9a sbi 0x12, 7 ; 18
d6: 97 98 cbi 0x12, 7 ; 18
if (DataMid & 0x10) SetDataHigh(); else SetDataLow(); SetClockHigh();
SetClockLow();
d8: 84 ff sbrs r24, 4
da: 35 c0 rjmp .+106 ; 0x146
dc: c0 9a sbi 0x18, 0 ; 24
de: 97 9a sbi 0x12, 7 ; 18
e0: 97 98 cbi 0x12, 7 ; 18
if (DataMid & 0x08) SetDataHigh(); else SetDataLow(); SetClockHigh();
SetClockLow();
e2: 83 ff sbrs r24, 3
e4: 2e c0 rjmp .+92 ; 0x142
e6: c0 9a sbi 0x18, 0 ; 24
e8: 97 9a sbi 0x12, 7 ; 18
ea: 97 98 cbi 0x12, 7 ; 18
if (DataMid & 0x04) SetDataHigh(); else SetDataLow(); SetClockHigh();
SetClockLow();
ec: 82 ff sbrs r24, 2
ee: 27 c0 rjmp .+78 ; 0x13e
f0: c0 9a sbi 0x18, 0 ; 24
f2: 97 9a sbi 0x12, 7 ; 18
f4: 97 98 cbi 0x12, 7 ; 18
if (DataMid & 0x02) SetDataHigh(); else SetDataLow(); SetClockHigh();
SetClockLow();
f6: 81 ff sbrs r24, 1
f8: 20 c0 rjmp .+64 ; 0x13a
fa: c0 9a sbi 0x18, 0 ; 24
fc: 97 9a sbi 0x12, 7 ; 18
fe: 97 98 cbi 0x12, 7 ; 18
if (DataMid & 0x01) SetDataHigh(); else SetDataLow(); SetClockHigh();
SetClockLow();
100: 80 ff sbrs r24, 0
102: 19 c0 rjmp .+50 ; 0x136
104: c0 9a sbi 0x18, 0 ; 24
106: 97 9a sbi 0x12, 7 ; 18
108: 97 98 cbi 0x12, 7 ; 18
if (DataLow & 0x04) SetDataHigh(); else SetDataLow(); SetClockHigh();
SetClockLow();
10a: 83 2f mov r24, r19
10c: 99 27 eor r25, r25
10e: 82 ff sbrs r24, 2
110: 10 c0 rjmp .+32 ; 0x132
112: c0 9a sbi 0x18, 0 ; 24
114: 97 9a sbi 0x12, 7 ; 18
116: 97 98 cbi 0x12, 7 ; 18
if (DataLow & 0x02) SetDataHigh(); else SetDataLow(); SetClockHigh();
SetClockLow();
118: 81 ff sbrs r24, 1
11a: 09 c0 rjmp .+18 ; 0x12e
11c: c0 9a sbi 0x18, 0 ; 24
11e: 97 9a sbi 0x12, 7 ; 18
120: 97 98 cbi 0x12, 7 ; 18
if (DataLow & 0x01) SetDataHigh(); else SetDataLow(); SetClockHigh();
SetClockLow();
122: 80 ff sbrs r24, 0
124: 02 c0 rjmp .+4 ; 0x12a
126: c0 9a sbi 0x18, 0 ; 24
128: 45 c0 rjmp .+138 ; 0x1b4
12a: c0 98 cbi 0x18, 0 ; 24
12c: 43 c0 rjmp .+134 ; 0x1b4
12e: c0 98 cbi 0x18, 0 ; 24
130: f6 cf rjmp .-20 ; 0x11e
132: c0 98 cbi 0x18, 0 ; 24
134: ef cf rjmp .-34 ; 0x114
136: c0 98 cbi 0x18, 0 ; 24
138: e6 cf rjmp .-52 ; 0x106
13a: c0 98 cbi 0x18, 0 ; 24
13c: df cf rjmp .-66 ; 0xfc
13e: c0 98 cbi 0x18, 0 ; 24
140: d8 cf rjmp .-80 ; 0xf2
142: c0 98 cbi 0x18, 0 ; 24
144: d1 cf rjmp .-94 ; 0xe8
146: c0 98 cbi 0x18, 0 ; 24
148: ca cf rjmp .-108 ; 0xde
14a: c0 98 cbi 0x18, 0 ; 24
14c: c3 cf rjmp .-122 ; 0xd4
14e: c0 98 cbi 0x18, 0 ; 24
150: bc cf rjmp .-136 ; 0xca
152: c0 98 cbi 0x18, 0 ; 24
154: b3 cf rjmp .-154 ; 0xbc
156: c0 98 cbi 0x18, 0 ; 24
158: ac cf rjmp .-168 ; 0xb2
15a: c0 98 cbi 0x18, 0 ; 24
15c: a5 cf rjmp .-182 ; 0xa8
15e: c0 98 cbi 0x18, 0 ; 24
160: 9e cf rjmp .-196 ; 0x9e
162: c0 98 cbi 0x18, 0 ; 24
164: 97 cf rjmp .-210 ; 0x94
166: c0 98 cbi 0x18, 0 ; 24
168: 90 cf rjmp .-224 ; 0x8a
16a: c0 98 cbi 0x18, 0 ; 24
16c: 89 cf rjmp .-238 ; 0x80
16e: c0 98 cbi 0x18, 0 ; 24
170: 82 cf rjmp .-252 ; 0x76
172: c0 98 cbi 0x18, 0 ; 24
174: 79 cf rjmp .-270 ; 0x68
176: 80 91 00 00 lds r24, 0x0000
17a: e8 2f mov r30, r24
17c: ff 27 eor r31, r31
17e: e0 50 subi r30, 0x00 ; 0
180: f0 40 sbci r31, 0x00 ; 0
182: c8 95 lpm
184: 90 2d mov r25, r0
186: 80 91 00 00 lds r24, 0x0000
18a: e8 2f mov r30, r24
18c: ff 27 eor r31, r31
18e: e0 50 subi r30, 0x00 ; 0
190: f0 40 sbci r31, 0x00 ; 0
192: c8 95 lpm
194: 20 2d mov r18, r0
196: 80 91 00 00 lds r24, 0x0000
19a: e8 2f mov r30, r24
19c: ff 27 eor r31, r31
19e: e0 50 subi r30, 0x00 ; 0
1a0: f0 40 sbci r31, 0x00 ; 0
1a2: c8 95 lpm
1a4: 30 2d mov r19, r0
1a6: 5c cf rjmp .-328 ; 0x60
1a8: 10 92 00 00 sts 0x0000, r1
1ac: 90 e8 ldi r25, 0x80 ; 128
1ae: 20 e3 ldi r18, 0x30 ; 48
1b0: 38 2f mov r19, r24
1b2: 56 cf rjmp .-340 ; 0x60
1b4: 97 9a sbi 0x12, 7 ; 18
1b6: 97 98 cbi 0x12, 7 ; 18
SetStrobeHigh(); // reset strobe to high to terminate
data transfer
1b8: 96 9a sbi 0x12, 6 ; 18
StepCompleted = TRUE;
1ba: 81 e0 ldi r24, 0x01 ; 1
1bc: 80 93 00 00 sts 0x0000, r24
}
1c0: ff 91 pop r31
1c2: ef 91 pop r30
1c4: 9f 91 pop r25
1c6: 8f 91 pop r24
1c8: 3f 91 pop r19
1ca: 2f 91 pop r18
1cc: 0f 90 pop r0
1ce: 0f be out 0x3f, r0 ; 63
1d0: 0f 90 pop r0
1d2: 1f 90 pop r1
1d4: 18 95 reti
000001d6 <__vector_8>:
and here is my makefile ...
CC=avr-gcc
OBJCOPY=avr-objcopy
OBJDUMP=avr-objdump
CFLAGS=-g -mmcu=at90s4433
RM = rm
SOURCES = demo.c
OBJECTS = demo.o
TARGETS = rom.hex demo.out
all : rom.hex steppr01.lst
rom.hex : steppr01.out
$(OBJCOPY) -j .text -O srec steppr01.out rom.hex
steppr01.out : steppr01.o
$(CC) $(CFLAGS) -o steppr01.out -Wl,-Map,steppr01.map steppr01.o -lm
# -lm above links in the math library needed for sqrt() fn.
steppr01.o : steppr01.c
$(CC) $(CFLAGS) -Os -c steppr01.c
steppr01.lst : steppr01.o
$(OBJDUMP) -S steppr01.o > steppr01.lst
clean:
-$(RM) $(OBJECTS)
realclean:
-$(RM) $(OBJECTS)
-$(RM) $(TARGETS)
Thanks in advance.
Rob Ward
- [avr-libc-dev] poor optimisation,
Rob Ward <=